I'm sorry, but I have to chime in here.
FILTERING NOTWITHSTANDING:
The digital data for a sine wave at exactly 1/2 of the sample rate (a
10k sine wave for 20k sampling) looks like this:
-_-_-_-_-_-_-_
The DATA will be:
-32768 ... 32767 ... -32768 ... 32767 ... etc.
it's EXACTLY THE SAME for a sine wave.
Now, as to if this makes a difference, if people can hear the
difference between a 20k sine and a 20k square, i couldn't say.
Imagine THIS, though. The digital data for a 7.5k anything (square,
sine, whatever) at 20k:
-_--__-_--__-_--__
at 20k, you CAN'T record a 7.5k sound, you can wiggle between 10k and
5k, though. Sure, my details on what it looks like might be wrong,
there are other ways it could be represented, but either way, it's
ugly.
The filter makes all this "OK" by reconstructing what the sound
should have been, by lopping off aliasing frequencies that
are 'outside human hearing.' This is also ugly.
Just try to imagine making something like this on graph paper, what
it would look like to try to represent something between the grid
lines.
So, really, I think you need at lesat 192k. Can you hear the
difference? Hell if I know, but at least the data is THERE.
Filtering off at 20k is just a hack. It's lame. I'd like to have a
system with NO filtering. Not likely, sure, but imagine a sampling
rate so high that at 20k, even drawing everything SQUARE your sound
would be so high rez that you couldn't tell. THAT's what i'm talking
about, yeah, yeah.
::nods off::
--- In motm@y..., Tim Walters <walters@d...> wrote:
> >That's already a given, and you missed my point anyway. If you
sample a
> >sine wave @ 20Khz and a square wave @ 20khz, you will only get a
10khz
> >square wave when you go D to A. The sine wave will lose detail.
>
> No, it won't. That's the whole point of the Nyquist theorem.
> Everything below the Nyquist frequency is reproduced ∗exactly∗
(given
> ideal filters etc.). A 20kHz sine wave is just as detailed when
> sampled at 44.1kHz as when sampled at 96kHz; either way, it
contains
> all the information of the original wave.
>
> The only thing increasing the sample rate does is allow you to
> represent higher frequencies (and possibly to design a better
> real-world filter).
>
> >I don't know what "statement" you're referring to, other than the
quality
> >of the waveform has zip to do with the Nyquist Theorem.
>
> That would be it.
> --
>
>
> -----------------------------------------------------------------
> Tim Walters : The Doubtful Palace : http://www.doubtfulpalace.com